MCQs

Disclaimer:

Educational Use Only:

The website links provided are to be used strictly for educational and knowledge-gaining purposes only.

No Misuse:

Do not misuse the links or the information contained therein for any commercial, illegal, or unethical activities.

Respect for Websites and Businesses:

Respect the intellectual property and business interests of the websites and their owners. Do not attempt to alter, copy, or redistribute content without proper authorization.

Acknowledge Value of Knowledge:

Recognize and value the knowledge and resources shared by these websites. Use the information responsibly and ethically.

No Endorsement:

The provision of these links does not imply endorsement of the linked sites, their content, or their products and services. Users are responsible for evaluating the suitability and accuracy of the information.

User Responsibility:

By accessing these links, you agree to use the information at your own risk and to comply with all applicable laws and website terms of use.

Overall MCQs for Reference

What is the main architectural innovation introduced by Transformers for handling sequential data?
a. Pooling layers
b. Self-attention mechanism
c. Convolutional filters
d. Recurrence
Answer: b. Self-attention mechanismgeeksforgeeks
Which component in Transformers is responsible for learning relationships between different tokens in a sequence?
a. Decoder
b. Feed-forward layer
c. Self-attention
d. Embedding layer
Answer: c. Self-attentiongeeksforgeeks+1
In Transformer architecture, what does positional encoding achieve?
a. Normalization
b. Sequence order awareness
c. Dropout
d. Feature scaling
Answer: b. Sequence order awarenessshiksha+1
Which is NOT an advantage of Transformer models over RNNs?
a. Better handling of long-range dependencies
b. Faster parallel computation
c. Limited to short texts
d. No vanishing gradient problem
Answer: c. Limited to short textsonlineexammaker+1
The model BERT stands for:
a. Bidirectional Encoder Recurrent Transformer
b. Bidirectional Encoder Representations from Transformers
c. Binary Encoder Recurrent Transformer
d. Binary Encoder Residual Transformer
Answer: b. Bidirectional Encoder Representations from Transformersgithub+1
What are the primary embeddings used in Transformer input?
a. Token, segment, position
b. Classification, pooling, convolution
c. Recurrent, attention, feed-forward
d. Word, sentence, paragraph
Answer: a. Token, segment, positionnaviera101+1
Which technique helps Transformers distinguish the position of each word in a sentence?
a. Dropout
b. Residual connection
c. Positional encoding
d. Layer normalization
Answer: c. Positional encodingshiksha+1
Which mechanism allows each token to focus on other tokens regardless of their position?
a. Feed-forward
b. Self-attention
c. Pooling
d. Recurrence
Answer: b. Self-attentiongeeksforgeeks+1
What type of layer is used after self-attention and before output in a Transformer block?
a. Convolutional layer
b. Feed-forward layer
c. Recurrent layer
d. Pooling layer
Answer: b. Feed-forward layeraiml+1
Transformers can be efficiently scaled due to which computation feature?
a. Parallelism in self-attention
b. Sequential recursion
c. Low memory requirement
d. Sparse connectivity
Answer: a. Parallelism in self-attentiononlineexammaker+1
The decoder in Transformer models is primarily used in which application?
a. Classification
b. Sequence generation
c. Object detection
d. Regression
Answer: b. Sequence generationonlineexammaker
What is a key limitation of Transformer models?
a. High computational cost
b. Poor text understanding
c. Limited scalability
d. Inefficient for long sequences
Answer: a. High computational costaiml+1
Which Transformer-based model is optimized for text classification?
a. GPT
b. ViT
c. BERT
d. ResNet
Answer: c. BERTgithub+1
Why do Transformers outperform traditional RNNs for NLP tasks?
a. They use pooling layers
b. They don’t need sequential computation
c. They use batch normalization
d. They depend on previous states
Answer: b. They don’t need sequential computationaiml+1
The process of fine-tuning a pre-trained Transformer model on a specific dataset is called:
a. Regularization
b. Transfer learning
c. Data augmentation
d. Embedding
Answer: b. Transfer learningaiml
Multi-head attention enables which of the following?
a. Pooling different outputs
b. Multiple attention perspectives
c. Faster dropout
d. Deeper layers only
Answer: b. Multiple attention perspectivesaiml
Which variant adapts Transformer architecture for image processing?
a. ImgNet
b. Vision Transformer (ViT)
c. ResNet
d. BERT
Answer: b. Vision Transformer (ViT)geeksforgeeks+1
What regularization technique is commonly used to improve Transformer training?
a. Batch normalization
b. Dropout
c. Max pooling
d. L1 regularization
Answer: b. Dropoutgeeksforgeeks+1
Transformers use which mechanism to update weights based on errors?
a. Momentum
b. Backpropagation
c. Adagrad
d. Forward pass
Answer: b. Backpropagationgeeksforgeeks+1
What does “masking” achieve in sequence-to-sequence Transformer models?
a. Data corruption
b. Prevents leaking future information
c. Random sampling
d. Reduces overfitting
Answer: b. Prevents leaking future informationgithub+1
In GPT models, which part of the Transformer is used for generating text?
a. Encoder
b. Decoder
c. Pooling layer
d. Embedding layer
Answer: b. Decoderonlineexammaker+1
Which framework provides major implementations of Transformer models?
a. TensorFlow
b. PyTorch
c. Both a and b
d. Scikit-learn
Answer: c. Both a and bgeeksforgeeks+1
Cross-attention in Transformers is mainly used for:
a. Handling multi-modal data
b. Focusing only on positional data
c. Dropping unrelated tokens
d. Bypassing self-attention
Answer: a. Handling multi-modal dataaiml
What is the primary use case of the decoder-only Transformer model?
a. Text understanding
b. Text generation
c. Image segmentation
d. Speech synthesis
Answer: b. Text generationonlineexammaker+1
For machine translation, Transformers achieve state-of-the-art by:
a. Using LSTM cells
b. Using self-attention and encoder-decoder
c. Using pooling layers
d. Using tree structures
Answer: b. Using self-attention and encoder-decoderonlineexammaker+1
What allows Transformers to process longer sequences than CNNs and RNNs?
a. Attention spans
b. Pooling
c. Residual connections
d. Positional encoding
Answer: d. Positional encodinggeeksforgeeks+1
Which optimization method is widely adopted for training Transformers?
a. Stochastic Gradient Descent
b. Adam
c. RMSProp
d. L-BFGS
Answer: b. Adamgeeksforgeeks+1
What is BERT mainly used for?
a. Image recognition
b. Audio synthesis
c. Text comprehension
d. Generative art
Answer: c. Text comprehensiongithub+1
Which Transformer variant is specialized for protein sequence analysis?
a. ProtBERT
b. BioGPT
c. SeqNet
d. AminoTransformer
Answer: a. ProtBERTaiml
What role does “layer normalization” play in Transformer models?
a. Prevents gradient explosion
b. Helps with convergence
c. Applies dropout
d. Optimizes embedding
Answer: b. Helps with convergencegeeksforgeeks+1
Which model improved on Transformer by introducing sparse attention mechanisms?
a. Longformer
b. BERT
c. GPT-2
d. ViT
Answer: a. Longformeraiml
The “attention is all you need” paper introduced which model?
a. LSTM
b. Transformer
c. CNN
d. RNN
Answer: b. Transformergeeksforgeeks+1
Which method is used for sequence-to-sequence tasks with Transformers?
a. Encoder-decoder
b. Convolutional stacking
c. Pooling
d. Recurrence
Answer: a. Encoder-decoderonlineexammaker+1
What increases the interpretability of Transformer model outputs?
a. Attention maps
b. Dropout rates
c. Residual connections
d. Pooling indices
Answer: a. Attention mapsgeeksforgeeks+1
How do pre-trained Transformer models improve downstream task performance?
a. By solving isolated tasks
b. By transferring prior knowledge
c. By removing irrelevant features
d. By using only labeled data
Answer: b. By transferring prior knowledgeshiksha+1
In Transformers, what does “multi-head” refer to?
a. Multiple output layers
b. Multiple attention mechanisms running in parallel
c. Multiple pooling strategies
d. Multiple recurrence loops
Answer: b. Multiple attention mechanisms running in parallelaiml
Which Transformer variant is designed for image classification?
a. ResNet
b. ViT
c. BERT
d. GPT-3
Answer: b. ViTgeeksforgeeks
What does masked language modeling in BERT involve?
a. Predict missing words
b. Ignore sentence order
c. Add random tokens
d. Drop tokens at random
Answer: a. Predict missing wordsgithub+1
Which parameter is not required in Transformer architecture?
a. Sequence length
b. Hidden size
c. Feature map
d. Input embedding
Answer: c. Feature mapaiml+1
What type of attention does a Transformer use?
a. Global self-attention
b. Single-head attention
c. Local recurrence
d. Pooling attention
Answer: a. Global self-attentiongeeksforgeeks+1
Training very large Transformer models may result in:
a. High generalization errors
b. Increased computational demand
c. Poor scalability
d. Lower accuracy
Answer: b. Increased computational demandshiksha+1
Which loss function is commonly used for training Transformers in language tasks?
a. Cross-entropy
b. Mean squared error
c. Hinge loss
d. Focal loss
Answer: a. Cross-entropyaiml
In machine translation, which mechanism allows Transformers to attend to source and target?
a. Pooling
b. Cross-attention
c. Feed-forward
d. Recurrence
Answer: b. Cross-attentionaiml
Which model is mainly used for conversational AI and chatbots?
a. GPT
b. BERT
c. ViT
d. ResNet
Answer: a. GPTjavaguides+1
Transformers reduce the vanishing gradient issue by:
a. Using skip connections
b. Using pooling
c. Limiting parameter size
d. Only training small models
Answer: a. Using skip connectionsgeeksforgeeks+1
How is inference time affected by Transformer model size?
a. Remains constant
b. Increases with size
c. Decreases with size
d. Completely independent
Answer: b. Increases with sizeshiksha+1
Which model architecture is NOT a Transformer variant?
a. XLNet
b. ResNet
c. BERT
d. GPT-3
Answer: b. ResNetaiml
What does the term “self-supervised pre-training” refer to in Transformers?
a. Training with labeled data
b. Training with unlabeled data and learning internal representations
c. Using only test data
d. Training with supervision
Answer: b. Training with unlabeled data and learning internal representationsshiksha+1
Which layer aggregates individual attention outputs in multi-head attention?
a. Output layer
b. Concat and linear
c. Pooling layer
d. Embedding projection
Answer: b. Concat and lineargeeksforgeeks+1
What was the primary impact of introducing the Transformer model for AI?
a. Improved image processing only
b. Breakthrough in NLP and generative modeling
c. Reduced text data requirements
d. Removed need for deep learning
Answer: b. Breakthrough in NLP and generative modelinggeeksforgeeks+1

Blog

Systemethic Technologies

Leave a Reply Cancel reply

Useful links

Other Terms & Policies

Contact Info